Find the general solution of $$2x\cos^2(y)+(x^2+1)\sin(y)y'=0.$$

First, I divided by $\sin(y)$: $$2x\frac{\cos^2(y)}{\sin(y)}+(x^2+1)y'=0\implies(x^2+1)y'=-2x\frac{\cos^2(y)}{\sin(y)},$$ so $$\frac{\sin(y)}{\cos^2(y)}y'=-\frac{2x}{x^2+1}\implies\int\frac{\sin(y)}{\cos^2(y)}\,\mathrm{d}y=\int-\frac{2x}{x^2+1}\,\mathrm{d}x.$$ For LHS: pick $u=\cos(y)$, so $\mathrm{d}u=-\sin(y)\,\mathrm{d}y$, then the integral becomes $$-\int\frac{1}{u^2}\,\mathrm{d}u=\frac{1}{u}+C_1=\frac{1}{\cos(y)}+C_1=\sec(y)+C_1.$$ For RHS: pick $u=x^2+1$, then $\mathrm{d}u=2x\,\mathrm{d}x$, so the integral becomes $$-\int\frac{1}{u}\,\mathrm{d}u=-\ln\lvert u\rvert+C_2=-\ln\lvert x^2+1\rvert+C_2\underbrace{=}_{x^2+1>0}-\ln(x^2+1)+C_2.$$ Finally, $$\sec(y)+C_1=-\ln(x^2+1)+C_2\implies\boxed{y=\sec^{-1}(-\ln(x^2+1)+C)}.$$ However, WA gives two solutions:


One of them is my solution, and the other is $y=-\sec^{-1}(-\ln(x^2+1)+C)$, which is also true because if we plug it into the differential equation we end up with $0$, and $0=0$ is true.

What did I miss?


The secant function is even (and not injective). This means that $$\sec(y)=f(x)$$ has multiple solutions $y$ for any particular value of $x$. Two of these solutions are $$y=\pm \sec^{-1}(f(x))$$

But there are even more solutions than that. Since $\sec(x)$ is periodic with period $2\pi$, we also have solutions like $$y=2\pi k\pm\sec^{-1}(f(x))$$ where $k\in\mathbb Z$.

  • $\begingroup$ Wow, you are right! Thanks! Since we have divided by $\sin(y)$ we cannot expect the values $k\pi$ ($k\in\Bbb{Z}$) for $y$, so the general solution is $y=\pm\sec^{-1}(-\ln(x^2+1)+C)$ but not $y=2\pi k\pm\sec^{-1}(-\ln(x^2+1)+C)$, right? $\endgroup$ – manooooh Jun 2 at 21:38
  • 1
    $\begingroup$ @manooooh Actually, $2\pi k \pm \sec^{-1}(-\ln(x^2+1)+C)$ solves the differential equation for any value of $k$. $\endgroup$ – Franklin Pezzuti Dyer Jun 2 at 21:40
  • $\begingroup$ Ok. Why WA does not show it? $\endgroup$ – manooooh Jun 2 at 21:42
  • $\begingroup$ @manooooh Beats me, but WA is certainly not infallible. $\endgroup$ – Franklin Pezzuti Dyer Jun 3 at 12:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.